A Knowledge-Recommendation Approach to Enhance Cognitive Efficacy in Engineering Design

The complexity and rapid updates of technical specification knowledge pose efficiency challenges for designers in extracting design specification knowledge. To address this issue, a knowledge recommendation method for engineering design based on user requirements and technical specification mapping is proposed. By analyzing the design specification text, the algorithm for extracting design knowledge identifies four types of knowledge: enumeration explanation, comparative relationship, conditional rule, and explanatory explanation. Additionally, a method is introduced to decompose natural language-expressed requirements into requirement units in engineering design. The development of a Neo4j-based algorithm has been undertaken to construct a visually represented knowledge graph for technical specifications. Taking the example of wind turbine related specification text in wind power generation field, a deep learning algorithm based on the BERT large language model is employed to align these findings with decomposed user requirements and provides designers with relevant technical specification knowledge. Ultimately, the proposed method demonstrates accurate recommendations of technical specification knowledge through validation with diverse user requirement cases. The average accuracy of the four types of rules is 99.66%, the average recall rate is 98.75%, and the average F1-score is 99.19%.